1: \documentclass[amsmath,twocolumn,superscriptaddress,showpacs,aps,prb]{revtex4}
2: \usepackage{bm}
3: \usepackage{graphics}
4: %\documentclass[aps,prl]{revtex4}
5: %\documentstyle{article}
6: %\textheight=240mm \textwidth=140mm \voffset=-30mm \hoffset=-20mm
7: %\usepackage{graphicx}
8: \begin{document}
9:
10: \title{Transport equations for a two-dimensional electron gas with spin-orbit interaction}
11: \author{E.G. Mishchenko}
12: %\email{also at L.D.\ Landau Institute}
13: \affiliation{Lyman
14: Laboratory, Department of Physics, Harvard University, MA 02138}
15: \author{B.I. Halperin}
16: \affiliation{Lyman Laboratory, Department of Physics, Harvard
17: University, MA 02138}
18:
19: \begin{abstract}
20: The transport equations for a two-dimensional electron gas with
21: spin-orbit interaction are presented. The distribution function is
22: a $2\times 2$-matrix in the spin space. Particle and energy
23: conservation laws determine the expressions for the electric
24: current and the energy flow. The derived transport equations are
25: applied to the spin-splitting of a wave packet and to the
26: calculation of the structure factor and the dynamic conductivity.
27: \end{abstract}
28:
29: \pacs{ 73.23.-b, 72.25.Dc, 73.21.-b}
30:
31: \maketitle Spin injection and coherent control of spins in
32: various nanostructures represent two principal challenges for the
33: field of spintronics. Recently, the amount of spintronics research
34: has grown up extensively with the ultimate goal of applications to
35: the quantum computing and information processing \cite{ALS}. A
36: number of spin-based devices have been designed and studied
37: \cite{DD,WAB,KNT,GBZ}. Spin manipulation in such devices can be
38: achieved by optical \cite{KA} or electric \cite{FKR,OYB,ZRK,RPS}
39: methods or by ferromagnetic gating \cite{fer}. A controlled
40: coupling between spin and orbital degrees of freedom is considered
41: to be a particularly promising tool of efficient spin manipulation
42: dating back to the seminal proposal by Datta and Das \cite{DD}.
43:
44: Spin-orbit interaction in two-dimensional electron gas (2DEG)
45: confined at GaAs/AlGaAs, GaN/AlGaN or similar heterojunctions
46: arises because of the quantum well asymmetry in the perpendicular
47: [$z$] direction. The resulting perpendicular electric field leads
48: to the coupling of spin to the electron momentum \cite{BR}. The
49: strength of this coupling can be experimentally tuned by a gate
50: voltage \cite{NAT,Koga}.
51:
52: Experimental advances in spin manipulation present a certain
53: challenge to develop a proper theoretical description for various
54: phenomena related to the spin-orbit interaction. In particular,
55: modification of universal conductance fluctuations and weak
56: localization has been studied in quantum dots \cite{H,AF,M}. The
57: phenomenon of weak localization has been considered in 2DEG as
58: well \cite{Koga,MZM}. The Friedel oscillations in the presence of
59: spin-orbit interaction\cite{CR} and the ac conductivity and the
60: plasmon attenuation\cite{MCE} are calculated.
61:
62: The principal goal of the present paper is to derive general
63: transport equations for the spin-dependent distribution function
64: of 2DEG including the effects of spin-orbital coupling. We assume
65: that the spin-orbit interaction in a two-dimensional electron gas
66: has the form,
67: \begin{equation}
68: \label{soham}
69: H_{so} =\alpha(\hat\sigma_x p_y-\hat\sigma_y p_x)
70: +\beta(\hat\sigma_x p_x-\hat\sigma_y p_y),
71: %+\gamma (\sigma_x p_x p_y^2 -\sigma_y p_y p_x^2)
72: \end{equation}
73: where the first term is the Bychkov-Rashba term \cite{BR} and the
74: second term is the linear Dresselhaus (or anisotropy) term present
75: in semiconductors with no bulk inversion symmetry \cite{D}. The
76: expression of Eq.\ (\ref{soham}) corresponds to the confinement
77: along the (001) growth direction \cite{DK}. Hereinafter we
78: neglect cubic Dresselhaus terms. The free particle Hamiltonian can
79: therefore be written in compact notations as,
80: \begin{equation}
81: \label{ham}
82: H= [\frac{{\bf p}^2}{2m}-\mu]\hat\sigma_0+ \alpha_{ik} \hat\sigma_i p_k, ~~ %i,k=x,y, ~~
83: ~\alpha_{ik} =\left(
84: \begin{array}{cc} \beta & \alpha \\ -\alpha & -\beta \end{array}
85: \right).
86: \end{equation}
87: Here, as usual, $\hat\sigma_0, \hat\sigma_x,\hat\sigma_y,
88: \hat\sigma_z$ constitute the set of Pauli matrices. We use Latin
89: subscripts for spatial coordinates and reserve Greek subscripts
90: for the spin indexes.
91:
92: {\it Spectral properties.} Before deriving the transport equation
93: we describe briefly the spectral properties of the Hamiltonian
94: (\ref{soham}). Its diagonalization is straightforward and reveals
95: the existence of two spin-split subbands in the electron
96: spectrum,
97: \begin{equation}
98: \label{spectrum} \epsilon_{1{\bf p}} = \xi_p+\Delta_{\bf p},~~~
99: \epsilon_{2{\bf p}} =\xi_p-\Delta_{\bf p},
100: \end{equation}
101: where
102: \begin{equation}
103: \label{spectrum1} \xi_p=\frac{p^2}{2m}-\mu, ~~~ \Delta_{\bf
104: p}=\sqrt{p^2(\alpha^2+\beta^2)+4\alpha\beta p_xp_y },
105: \end{equation}
106: and $p^2=p_x^2+p_y^2$ is the total electron momentum. The
107: eigenfunctions corresponding to the eigenstates (\ref{spectrum})
108: are,
109: \begin{equation}
110: \label{states} \psi_{1,2} ({\bf x}) =\frac{1}{\sqrt{2}} \left(
111: \begin{array}{c} e^{i\chi_{\bf p}/2}
112: \\ \pm e^{-i\chi_{\bf p}/2}
113: \end{array} \right) e^{i{\bf p x}},
114: %~~~\psi_2(x) =\frac{1}{\sqrt{2}} \left( \begin{array}{c} e^{i\chi_{\bf p}/2} \\
115: %-e^{-i\chi_{\bf p}/2}
116: %\end{array} \right) e^{i{\bf p x}},
117: \end{equation}
118: with the upper (lower) sign corresponding to the $\psi_1$
119: ($\psi_2$) state. The phase factor $\chi_{\bf p}$ depends on the
120: direction of the electron momentum,
121: \begin{equation}
122: \tan \chi_{\bf p} = \frac{\alpha p_x+\beta p_y}{\alpha p_y +\beta
123: p_x}.
124: \end{equation}
125: For the isotropic 2DEG ($\beta =0$) the phase $\chi_{\bf p}$
126: coincides with the angle between the electron momentum and the
127: $y$-axis.
128:
129: Further and more convenient description of the spectral properties
130: can be obtained by considering the spin-dependent retarded Green
131: function, defined as usual by,
132: \begin{equation}
133: \label{gr} i G^R_{\alpha \beta}(x,x') =\theta(t-t') \langle
134: \langle \psi_\alpha (x)\psi_\beta^\dagger (x')+\psi_\beta^\dagger
135: (x') \psi_\alpha (x) \rangle \rangle.
136: \end{equation}
137: Here we have used the shorthand notation for the space and time
138: variables, $x=({\bf x},t)$. In a homogeneous system the
139: correlation functions depend on the relative coordinates $x-x'$
140: only. Using the above expressions (\ref{spectrum}-\ref{states}) we
141: can write the expression for the retarded Green function of free
142: electrons in the momentum representation, $\hat G^R (\epsilon,{\bf
143: p}) =\int dt d{\bf x} ~\hat G^R (x) e^{i\epsilon t-i{\bf p}{\bf
144: x}}$, which after simple transformations takes the form,
145: \begin{eqnarray}
146: \label{greq} \hat G^R(\epsilon,{\bf p})&=& \sum_{\mu=1,2}\hat
147: G_\mu^R(\epsilon,{\bf p}) = \frac{1}{2} \frac{\hat
148: \sigma_0+\cos{\chi_{\bf p}}\hat\sigma_x
149: -\sin{\chi_{\bf p}}\hat \sigma_y}{\epsilon -\epsilon_{1{\bf p}}+i\eta}\nonumber\\
150: && +\frac{1}{2} \frac{\hat \sigma_0-\cos{\chi_{\bf p}}\hat\sigma_x
151: +\sin{\chi_{\bf p}}\hat \sigma_y}{\epsilon -\epsilon_{2{\bf
152: p}}+i\eta},
153: \end{eqnarray}
154: with the indexes $\mu=1,2$ corresponding to the first and second
155: terms, respectively, in the last expression of Eq.\ (\ref{greq}).
156:
157: The central quantity in the transport theory is the density
158: matrix,
159: \begin{equation}
160: \label{densitym} f_{\alpha \beta}(x,x') = \langle \langle
161: \psi_\beta^\dagger (x') \psi_\alpha (x) \rangle \rangle.
162: \end{equation}
163: Its value in the thermal equilibrium is related to the imaginary
164: part of the retarded Green function via the
165: fluctuation-dissipation theorem,
166: \begin{equation}
167: \label{fdt} \hat f(\epsilon,{\bf p}) = \sum_{\mu=1,2} n_{\mu\bf p}
168: ~[ \hat G^{R\dagger}_\mu (\epsilon,{\bf p})-G^{R}_\mu
169: (\epsilon,{\bf p})],
170: \end{equation}
171: here $n_{\mu \bf p}$ is the Fermi-Dirac distribution function for
172: the $\mu$-th state (\ref{spectrum}). It is convenient to expand
173: the density matrix over the complete set of the Pauli matrices. In
174: particular, for the equal-time density matrix,
175: \begin{equation}
176: \label{pauli} \hat f_{\bf p} = \int \frac{d\epsilon}{2\pi} \hat
177: f(\epsilon,{\bf p})= \frac{1}{2} f_{\bf p}\hat\sigma_0+\frac{1}{2}
178: {\bf g}_{\bf p}\cdot \hat{\bm{\sigma}},
179: \end{equation}
180: we observe according to Eq.\ (\ref{fdt}) that in the thermal
181: equilibrium,
182: \begin{equation}
183: \label{equil}
184: \begin{array}{ll} f_{\bf p}=(n_{1\bf p}+n_{2\bf p}),& ~g_{{\bf p}x}=\cos\chi_{\bf p}
185: (n_{1\bf p}-n_{2\bf p}),\\ g_{{\bf p}z}=0, & ~g_{{\bf
186: p}y}=-\sin\chi_{\bf p}(n_{1\bf p}-n_{2\bf p}),
187: \end{array}
188: \end{equation}
189:
190: {\it Transport equations}. In a generic nonequilibrium state, the
191: density matrix (\ref{densitym}) obeys a set of conjugated
192: equations that can be obtained from the equations of motion for
193: the electron operators $\psi(x)$ and $\psi^\dagger(x)$ determined
194: by the Hamiltonian (\ref{ham}),
195: \begin{eqnarray}
196: \label{set} \left[ i\partial_t +\frac{\nabla^2}{2m}+\mu
197: -e\phi_x\right]\hat f(x,x')
198: +i\alpha_{ik} \hat\sigma_i \nabla_k \hat f(x,x')=0,\nonumber\\
199: \left[i\partial_{t'} -\frac{\nabla'^2}{2m}-\mu
200: +e\phi_{x'}\right]\hat f(x,x') +i\alpha_{ik} \nabla_k' \hat
201: f(x,x')\hat \sigma_i =0.\nonumber\\
202: \end{eqnarray}
203: The equations (\ref{set}) neglect impurity scattering. This is
204: justified for ballistic systems when the mean free path exceeds
205: the characteristic system size, e.g.\ in high-mobility 2DEG in
206: semiconductor heterostructures (we discuss the impurity scattering
207: at the end of the paper). In Eq.\ (\ref{set}) we allowed for the
208: scalar external field $\phi_x=\phi({\bf x},t)$. In the absence of
209: electron-impurity or electron-electron collisions no self-energy
210: terms appear in the equations (\ref{set}) which makes it
211: sufficient to consider the equal-time ($t=t'$) functions only.
212:
213: Following the known route of deriving kinetic equations \cite{LP}
214: we utilize the Wigner transformation for the density matrix,
215: \begin{equation}
216: \label{wign} \hat f_{\bf p}({\bf x}, t)= \int d {\bf r}~ e^{-i \bf
217: pr} \hat f({\bf x}+\frac{\bf r}{2}, {\bf x}-\frac{\bf r}{2},t ).
218: \end{equation}
219: By taking the sum of equations (\ref{set}) in the Wigner
220: representation we obtain,
221: \begin{eqnarray}
222: \label{tr} [ \partial_t +{\bf v}\cdot\nabla ] \hat f_{\bf
223: p}+ie\int d{\bf q}~\phi_{\bf q}
224: (\hat f_{{\bf p}-\frac{\bf q}{2}}-\hat f_{{\bf p}+\frac{\bf q}{2}})e^{i{\bf qx}}\nonumber \\
225: + i\alpha_{ik} p_k[\hat\sigma_i,\hat f_{\bf p}]+\frac{1}{2}
226: \alpha_{ik}\nabla_k \{\hat\sigma_i, \hat f_{\bf p}\}=0,
227: \end{eqnarray}
228: where ${\bf v}={\bf p}/m$. Here we introduced the spatial Fourier
229: transform for the scalar potential $\phi_{\bf q}= \int d{\bf x}
230: \phi({\bf x},t)e^{i{\bf qx}}$, with the shorthand notation for the
231: momentum integration $d{\bf q}=d^2q/(2\pi)^2$.
232:
233: Finally, to present Eq. (\ref{tr}) in a more transparent way we
234: turn to the Pauli matrix representation (\ref{pauli}) to write,
235: \begin{eqnarray}
236: \label{tr1} [ \partial_t +{\bf v}\cdot\nabla]~ f_{\bf p}&+&ie\int
237: d{\bf q}~\phi_{\bf q} (f_{{\bf p}-\frac{\bf q}{2}}-f_{{\bf
238: p}+\frac{\bf q}{2}})e^{i{\bf qx}}\nonumber \\ &+& \alpha_{ik}
239: \nabla_k g_{{ \bf p}i} =0,
240: \end{eqnarray}
241: \begin{eqnarray}
242: \label{tr2} [ \partial_t +{\bf v}\cdot\nabla] ~ g_{{\bf p }i}
243: &+&ie\int d{\bf q}~\phi_{\bf q} (g_{{\bf p}-\frac{\bf
244: q}{2}i}-g_{{\bf p}+\frac{\bf q}{2}i})e^{i{\bf qx}}\nonumber\\
245: \nonumber\\ &-&[{\bf b}_{\bf p}\times {\bf g}_{\bf p}]_i
246: +\alpha_{ik} \nabla_k f_{\bf p}=0,
247: \end{eqnarray}
248: with the following notation for the precession frequency, ${ b}_{
249: {\bf p}i}= 2\alpha_{ik}p_k = 2\Delta_{\bf p} (\cos\chi_{\bf
250: p},-\sin{\chi_{\bf p}},0)$.
251:
252: {\it Conservation laws.} The transport equations
253: (\ref{tr1},\ref{tr2}) are of the Boltzmann type and therefore
254: fulfill certain particle and energy conservation conditions which
255: will now be obtained. By integrating Eq. (\ref{tr1}) with respect
256: to the momentum we find the continuity equation for the particle
257: flow,
258: \begin{equation}
259: \label{cont} \partial_t \rho +\frac{1}{e}\nabla \cdot {\bf j} =0,
260: \end{equation}
261: where the electron density and the electric current are given
262: respectively by,
263: \begin{eqnarray}
264: \label{density}
265: \rho = \int d{\bf p} f_{\bf p},~~~ j_k = e\int
266: d{\bf p}~ [v_k f_{\bf p}+ \alpha_{ik} g_{{\bf p}i}].
267: \end{eqnarray}
268: %Here $d{\bf p}=d^2p/(2\pi)^2$ stands for the element in the
269: %momentum space.
270: The terms containing the external potential $\phi_{\bf q}$ cancel
271: as is readily seen by the change of integration variables. To
272: obtain the energy continuity condition we multiply Eq. (\ref{tr1})
273: by $\xi_p$ and Eq. (\ref{tr2}) by ${\bf b}_{\bf p}$ and add them
274: together. After simple transformations the conservation of energy
275: can be written in the conventional form,
276: \begin{equation}
277: \label{heat}
278: \partial_t \rho^\epsilon +\nabla \cdot {\bf j}^\epsilon ={\bf j}\cdot {\bf E},
279: \end{equation}
280: where the energy density and energy current are,
281: \begin{eqnarray*}
282: \rho^\epsilon &=& \int d{\bf p} [\xi_p f_{\bf p}+ {b}_{{\bf
283: p}i} {g}_{{\bf p}i}],\nonumber\\
284: j^\epsilon_k &=& \int d{\bf p} v_k[\xi_p f_{\bf p}+ {b}_{{\bf
285: p}i}g_{{\bf p}i}]+\alpha_{ik} \int d{\bf p}[\xi_p g_{{\bf p}i}
286: +b_{{\bf p}i} f_{\bf p}].
287: \end{eqnarray*}
288: The equation (\ref{heat}) means that the local energy change is
289: due to the energy flow to the neighboring points in space as well
290: as a result of the local Joule heating (right-hand side).
291:
292: {\it Wave packet splitting.} To give a specific application of the
293: derived equations let us now use them to describe the propagation
294: of a wave packet in 2DEG with a spin-orbit coupling. We neglect a
295: spin-orbit anisotropy $\beta=0$ for simplicity. The wave packet
296: propagates along the $y$-direction and is uniform along the
297: $x$-axis. The transport equations (\ref{tr1},\ref{tr2}) are then
298: one-dimensional and (with no external field applied) yield,
299: \begin{eqnarray}
300: \label{packet}
301: &&\left[ \partial_t +v\partial_y \right] f=-\alpha \partial_y g_x,\nonumber\\
302: && \left[ \partial_t + v\partial_y\right]g_x=-\alpha \partial_y
303: f,\nonumber\\
304: &&\left[ \partial_t + v\partial_y\right]g_y=-2\Delta_p g_z
305: \nonumber\\
306: && \left[
307: \partial_t + v\partial_y\right]g_z= 2\Delta_p g_y.
308: \end{eqnarray}
309: First, we consider a spin-unpolarized Gaussian wave packet
310: injected at the point $y=0$ at the time $t=0$ and moving with the
311: average momentum $\bar p$,
312: \begin{equation}
313: \label{fp} \hat f_{\bf p}({\bf x},t=0) = \hat \sigma_0 F (y), ~~
314: F(y)=e^{-y^2\delta p^2 -\frac{(p_y-\bar p)^2}{\delta p^2}}.
315: \end{equation}
316: In this geometry the phase factor $\chi_{\bf p}=0$, which means
317: that the precession vector ${\bf b}$ is directed along the
318: $x$-axis. We also observe that $g_y=g_z=0$. The remaining two of
319: the equations (\ref{packet}) are easily solved by Fourier
320: transforming them into a set of linear algebraic equations.
321: % By
322: %Fourier transforming these equations,
323: %\begin{eqnarray*}
324: %(\omega-kv) f(\omega,k) =\alpha k g_x(\omega,k), \\
325: %(\omega-kv) g_x(\omega,k) =\alpha k f(\omega,k),
326: %\end{eqnarray*}
327: %it is straightforfard to see that their solution can be written as,
328: %\begin{eqnarray*}
329: %f(\omega,k) = A(k) \delta (\omega-kv-k\alpha)+B(k) \delta (\omega-kv+k\alpha),\\
330: %g_x(\omega,k) = A(k) \delta (\omega-kv-k\alpha)-B(k) \delta (\omega-kv+k\alpha),
331: %\end{eqnarray*}
332: %which means that in the real space and time,
333: The general solution of Eqs. (\ref{packet}) takes the form,
334: \begin{eqnarray*}
335: f(y,t) = A(y-v_+t)+B(y-v_-t),\\
336: g_x(y,t) = A(y-v_+t)-B(y-v_-t),
337: \end{eqnarray*}
338: where we have introduced subband velocities $v_\pm = v \pm
339: \alpha$. So far, $A(x)$ and $B(x)$ are two arbitrary functions
340: which have to be determined from the initial condition
341: (\ref{packet}) yielding, $ A(y)=B(y) = F(y) $. We find that the
342: incident wave packet (\ref{fp}) is decomposed into two independent
343: constituents oppositely polarized along $x$-direction and moving
344: with different velocities. The spatial distribution of the
345: electron density is given by the integral over all momenta, i.e.
346: \begin{eqnarray}
347: \label{split} \rho_\pm (y,t) = \frac{1}{2\pi^{1/2} \delta x (t)}
348: \exp{[-\frac{(x-\bar v_\pm t)^2}{\delta x^2 (t)}]},
349: \end{eqnarray}
350: with the average velocities $\bar v_\pm =\bar p/m \pm \alpha$, and
351: the Gaussian width at finite times, $\delta x^2 (t) =\delta
352: p^{-2}+\frac{t^2\delta p^2}{m^2}$. To observe the spin-orbit
353: induced splitting of a wave packet the following conditions should
354: be satisfied,
355: $$
356: \frac{\delta p}{m} \ll \alpha \ll \frac{t}{\delta p}.
357: $$
358: The first of the two conditions ensures that the splitting
359: dominates over the wave packet broadening, while the second
360: condition means that enough time has to elapse before the
361: splitting becomes larger than the intrinsic packet width.
362:
363: Now let us consider an injection of a packet initially polarized
364: along the $y$-direction.
365: \begin{equation}
366: \label{fp2} \hat f_{\bf p}({\bf x}) = (\hat \sigma_0 +\hat
367: \sigma_y) F (y),
368: \end{equation}
369: The equations for the $f$ and $g_x$ components of the density
370: matrix remain unchanged with the above analysis still valid. The
371: second pair of Eqs. (\ref{packet}) is independent of the first
372: pair and have a solution,
373: \begin{eqnarray}
374: \label{precession}
375: g_y(y,t) = F(y-vt)\cos{(2\Delta_pt)},\nonumber\\
376: g_z(y,t) = -F(y-vt)\sin{(2\Delta_pt)}.
377: \end{eqnarray}
378: According to the expressions (\ref{precession}) the initial spin
379: polarization precesses with a frequency $2\Delta_p$ around the
380: axis perpendicular to the propagation direction. Note that the
381: precessing spin propagates with the center-of-mass velocity $\bar
382: v$ rather than with the subband velocities $\bar v_\pm$.
383:
384: The above analysis assumes that a wavepacket is injected with a
385: given momentum $\bar p$. Such an injection into 2DEG with a
386: spin-orbit coupling is not easy to achieve. For example, injection
387: through an interface with a 'normal' (with no spin-orbit
388: interaction) 2DEG\cite{HM,MH,LLF} would not result in a spatial
389: splitting of a wave packet. This is due to the fact that the
390: injection happens with a conservation of energy rather than
391: momentum. As seen from Eqs. (\ref{spectrum}-\ref{spectrum1}) the
392: two states with the same energy propagate with the same
393: velocity\cite{MSB,LLF} within the approximations of this paper.
394: However, if we take into account the cubic Dresselhaus terms,
395: which have been omitted in our discussion, there can be a
396: splitting of velocities at the same energy. In order to achieve
397: splitting without the cubic terms we need to consider a more
398: complicated setup. As a demonstration of principle, we consider
399: the following example. Let us inject a wave packet propagating
400: along the $y$-direction with the spin polarized along the
401: interface ($x$-axis), e.g. by injection from a ferromagnetic
402: contact. The states forming the wavepacket belong to the subband
403: 1, with a spin polarization $\frac{1}{\sqrt 2} (1,1)$. Let us now
404: switch on ac magnetic field along the $y$-axis rotating the spin
405: direction until it is aligned with the $z$-axis, $(1,0)$, and then
406: switch the magnetic field off. The resulting state will be an
407: equal mixture of both eigenstates $\frac{1}{\sqrt 2} (1,1)$ and
408: $\frac{1}{\sqrt 2} (1,-1)$ without any change of momentum (the
409: energy is no longer conserved). The velocities of these states are
410: different and the packet will split.
411:
412: The above picture holds not only for the injection of initially
413: polarized packet. If the incident packet is unpolarized and has a
414: given energy, upon entering the interface it will become a mixture
415: of two states: $\frac{1}{\sqrt 2} (1,1)$ with the momentum
416: $p_0-m\alpha$, and $\frac{1}{\sqrt 2} (1,-1)$ with the momentum
417: $p_0+m\alpha$. Both velocities remain equal to $v_0=p_0/m$. After
418: switching on the ac magnetic field with the frequency $\omega
419: \approx 2\alpha p_0$ (which is a resonant frequency for the
420: transition between the two subbands), the first state will evolve
421: into the mixture of the states: $\frac{1}{\sqrt 2} (1,1)$ and
422: $\frac{1}{\sqrt 2} (1,-1)$, both with the momentum $p_0+m\alpha$,
423: meaning two different velocities $v_0$ and $v_0-2\alpha$. The same
424: reasoning shows that the other initial state will develop two
425: velocities $v_0$ and $v_0+2\alpha$. Therefore, the initially
426: unpolarized packet will split into three parts.
427:
428: {\it Ballistic spin injection}. We envisage a spin injection from
429: ferromagnetic contacts into ballistic 2DEG among the applications
430: for the equations derived above. In this case the injection occurs
431: with conservation of energy, and can be described by the
432: time-independent solution of the equations (\ref{tr1}-\ref{tr2})
433: with the appropriate boundary conditions, which require a
434: conservation of the normal components of the electric current,
435: Eq.\ (\ref{density}), at the interfaces. A corresponding theory
436: would generalize the existing approach for the ballistic
437: spin-injection based on the ordinary Boltzmann
438: equation\cite{Rashba}. In the latter case the Boltzmann equation
439: method is more convenient for the calculation of spin polarization
440: of current and magnetoresistance than the direct solution of the
441: single-particle Schr\"odinger equation.
442:
443: {\it Structure factor.} The electron density
444: fluctuations are described by the structure factor \cite{PN}
445: defined as the retarded correlation function,
446: \begin{eqnarray}
447: \label{struc} \chi(x,x') = -i\theta(t-t') \langle \langle
448: \rho(x)\rho (x')-\rho (x') \rho (x) \rangle \rangle,
449: \end{eqnarray}
450: of the electron density operators $\rho (x)= \psi_\alpha^\dagger
451: (x) \psi_\alpha (x)$. At equilibrium the structure factor
452: (\ref{struc}) depends on the relative coordinates $x-x'$ only. The
453: imaginary part of the Fourier transform $\chi (\omega,{\bf q})$
454: measures the energy dissipation of the external field at a given
455: frequency $\omega$ and a wavevector ${\bf q}$. In the isotropic
456: system the structure factor is related to the ac conductivity by
457: the relation,
458: \begin{eqnarray} \label{conduc} \sigma (\omega)
459: =\lim_{q \to 0} \frac{ie^2\omega}{q^2} \chi (\omega,{\bf q}).
460: \end{eqnarray}
461: The formula (\ref{conduc}) is readily checked using the Kubo
462: formula for the conductivity and the continuity equation
463: (\ref{cont}).
464: %In the isotropic system according to Eq.
465: %(\ref{conduc}) the structure factor completely determines the
466: %conductivity. The imaginary part of the structure factor measures
467: %therefore the energy dissipation of the external field at a given
468: %frequency $\omega$ and a wavevector ${\bf q}$.
469:
470: According to the fluctuation-dissipation theorem, the structure
471: factor can be determined from calculations of the linear response
472: to an
473: external scalar field. The field-induced modulation of electron
474: density is related to the magnitude of the external perturbation
475: through the structure factor according to \cite{PN},
476: \begin{eqnarray}
477: \label{relation} \delta \rho (\omega,{\bf q}) = e \chi(\omega,{\bf
478: q}) \phi (\omega,{\bf q}).
479: \end{eqnarray}
480: The electron density modulation is given by the deviation of the
481: function $f_{\bf p}(t,{\bf x})$ from its equilibrium value,
482: $\delta \rho (\omega,{\bf q }) = \int d{\bf p} ~\delta f_{\bf p}
483: (\omega,{\bf q }),$ and can be found from the linearized equations
484: (\ref{tr1},\ref{tr2}).
485: In the linear approximation by the external
486: field $\phi(\omega, {\bf q})$, the distribution function is a
487: small deviation
488: \begin{eqnarray}
489: \label{delta}
490: f_{\bf p} =f^0_{\bf p} +\delta f_{\bf p},~~~
491: {\bf g}_{\bf p} = {\bf g}^0_{\bf p}+\delta {\bf g}_{\bf p},
492: \end{eqnarray}
493: from its equilibrium value (\ref{equil}). The linearized transport
494: equations (\ref{tr1},\ref{tr2}) take the form,
495: \begin{eqnarray}
496: \label{linear} (\omega -{\bf qv} )\delta f_{\bf p}&-&\alpha_{ik}
497: q_i \delta {g}_{{\bf p}k}= e\phi (\omega, {\bf
498: q}) (f^0_{{\bf p}+\frac{\bf q}{2}}+f^0_{{\bf p}-\frac{\bf q}{2}}),\nonumber \\
499: (\omega -{\bf qv} ) \delta {g}_{{\bf p}i}&-&i [{\bf b}_{\bf p}
500: \times \delta {\bf g}_{\bf p}]_i -\alpha_{ik} q_k \delta f_{\bf p} = \nonumber\\
501: &-& e \phi (\omega, {\bf
502: q}) ({g}^0_{{\bf p}+\frac{\bf q}{2}i}+{\bf g}^0_{{\bf p}-\frac{\bf
503: q}{2}i}).
504: \end{eqnarray}
505: Solving these equations for the variation of the electron density
506: (\ref{density}) we obtain the structure factor with the help of
507: the relation (\ref{relation}),
508: %\begin{widetext}
509: \begin{eqnarray}
510: \label{chi} \chi(\omega, {\bf q}) &=& \frac{1}{2} \sum_{\mu \mu'}
511: \int d{\bf p} ~ [1+(-1)^{\mu\mu'}\cos{(\chi_{{\bf p}}-\chi_{{\bf
512: p'}})}] \nonumber\\ && \times \frac{n_{\mu {\bf p}_-}-n_{\mu'{\bf
513: p}_+}}{\omega-\epsilon_{\mu'{\bf p}_+}+\epsilon_{\mu {\bf p}_-}},
514: \end{eqnarray}
515: %\end{widetext}
516: where ${\bf p}_\pm={\bf p}\pm {\bf q}/2$. The expression
517: (\ref{chi}) with $\omega=0$ corresponds to the previously derived result
518: for the static dielectric
519: function\cite{CR}.
520: To simplify further the subsequent discussion we
521: will disregard the anisotropy, $\beta =0$, and consider the
522: zero-temperature limit $T=0$. The two spin-orbit subbands are
523: axially symmetric, shown on Fig. 1. The subbands are filled up to
524: the same Fermi energy level $\epsilon_F$ but have two different
525: Fermi momenta, $p_1$ and $p_2$, determined from the equations
526: $\epsilon_i(p_i)= \epsilon_F$, where $\epsilon_i(p)$ are given by
527: Eqs. (\ref{spectrum},\ref{spectrum1}) with $\beta =0$. This leads
528: to the values,
529: \begin{eqnarray}
530: p_1= p_0-m\alpha +O(m^2\alpha^2/p_0^2),\nonumber\\
531: p_2= p_0+m\alpha +O(m^2\alpha^2/p_0^2),
532: \end{eqnarray}
533: where $p_0$ is determined by $\epsilon_F=p_0^2/2m$, namely $p_0$
534: is the Fermi-momentum in the absence of spin-orbit interaction.
535: Note that the Fermi velocities for the two subbands,
536: \begin{equation}
537: v_i = \frac{\partial \epsilon_i(p)}{\partial p} \vert_{p=p_i} =
538: \frac{p_0}{m}+O(m^2\alpha^2/p_0^2),
539: \end{equation}
540: are the same and (up to higher order terms) equal to the Fermi
541: velocity in 2DEG with no spin-orbit coupling $\alpha =0$.
542: \begin{figure}
543: \resizebox{.33\textwidth}{!}{\includegraphics{bands.eps}}
544: \caption{Spin-orbit induced subbands of an isotropic
545: two-dimensional electron gas, $\epsilon_p=p^2/2m \pm \alpha |p|$.
546: At $T=0$ all states below the Fermi energy $\epsilon_F$ are
547: filled. The Fermi momenta for the two subbands are $p_{1,2}=p_0\mp
548: m\alpha$. The direct transitions, $q=0$ (shown by the arrow) are
549: possible for the states between the dashed lines, $p_1 < p <p_2$.}
550: \end{figure}
551: The imaginary part of the structure factor $\chi(\omega,{\bf q})$
552: determines the absorption, or Landau damping, of the external
553: field at given frequency and wavevector. The points in the
554: electron momentum space that contribute to the Landau damping
555: correspond to the zeros of the denominators. There are total four
556: determined by the equations,
557: \begin{equation}
558: \label{excit} \omega ={\bf qv} \pm \alpha {p}_+\pm \alpha p_-,
559: \end{equation}
560: with the opposite signs of the last two terms corresponding to the
561: (gapless) transitions within the same subbands [Eq.\
562: (\ref{spectrum})] and equal signs describing the transitions
563: between different subbands.
564:
565: The terms with $\mu=\mu'$ in Eq. (\ref{chi}) represent the effect of
566: intrasubband transitions. Only indirect ($q \ne 0$) transitions
567: contribute to the imaginary part of the structure factor. For
568: small transferred momenta $q \ll p_0$ one can disregard the
569: deviation of the cosine factor from unity and also approximate
570: $n_{i{\bf p}_-}-n_{i{\bf p}_+} \simeq - {\bf q} \partial n_{i}
571: /\partial {\bf p} $. Taking the momentum integral we obtain for
572: the contribution of the $i$-th subband
573: \begin{eqnarray}
574: \label{Im} \Im \chi_i (\omega,q)&=& -\nu_i \frac{
575: \omega}{\sqrt{q^2v_0^2 -\omega^2 }}~\theta(q^2v^2_0-\omega^2),
576: \end{eqnarray}
577: where $v_0=p_0/m$ and $\nu_i$ stands for the density of states of
578: the $i$-th subband at its Fermi surface $p=p_i$:
579: $\nu_1=\frac{m}{2\pi} (1-\frac{m\alpha}{p_0})$,
580: $\nu_2=\frac{m}{2\pi} (1+\frac{m\alpha}{p_0})$. Note that the sum
581: of the two contributions (\ref{Im}) is {\it independent} of the
582: spin-orbit interaction (up to higher-order terms), a consequence
583: of the fact that the two subbands have the same value of the
584: Fermi-velocity. Spin-orbit interaction results only in a
585: redistribution of the spectral weight between the subbands
586: controlled by the changes in the densities of states.
587:
588: The terms with $\mu\ne \mu'$ in Eq. (\ref{chi}) correspond to the
589: intersubband transitions. Their contribution to the structure
590: factor for $m\alpha \ll q \ll p_0$ is negligible compared to the
591: above considered intrasubband transitions by the factor $\sim
592: q^2/p_0^2$ (due to the small $\sin^2$ prefactor). However, the
593: presence of the two subbands is important as it makes the direct,
594: $q=0$, transitions possible. The factor $n_{1p}-n_{2p}$ then
595: defines the momentum space available for the direct transitions, $
596: p_1 < p <p_2$ (see Fig. 1), which corresponds to the frequency
597: domain $2\Delta_0 - 2m\alpha^2 <\omega < 2\Delta_0 +2m\alpha^2$,
598: where $\Delta_0= \Delta_{p_0}$,
599: \begin{equation}
600: \label{direct} \chi (\omega,q \to 0) = \frac{\alpha q^2}{4\pi}
601: \int\limits_{p_1}^{p_2} \frac{dp}{(\omega+i0)^2-4\Delta_p^2}.
602: \end{equation}
603: The imaginary part of this expression is
604: \begin{equation}
605: \label{imdirect} \Im \chi(\omega,q \to 0) = -\frac{q^2 \text{sgn}~
606: \omega }{32\Delta_0} ~ \theta [4m^2\alpha^4 -(\omega-
607: 2\Delta_0)^2].
608: \end{equation}
609: The equation (\ref{direct}) corresponds to the previously obtained
610: result \cite{MCE} for the optical conductivity $\sigma (\omega)$.
611: The expression (\ref{direct}) goes to zero with the wavevector,
612: which is easily understood by noting that the matrix elements for
613: the transitions between $\psi_1({\bf p}_-) $ and $\psi_{2}({\bf p
614: }_+)$ states are suppressed at small transferred momenta since
615: they are orthogonal at $q=0$. However, their contribution to the
616: conductivity [according to Eq. (\ref{conduc})] remains finite,
617: which is clear since the operator of electron velocity has nonzero
618: matrix elements for the intersubband transitions even at $q=0$.
619:
620: The experimental observation of the direct transitions
621: (\ref{imdirect}) is feasible in the measurements of the resonant
622: microwave absorption in high-mobility semiconductor
623: heterostructures.
624:
625: {\it Screened electron-electron interaction and plasmon
626: excitations.} So far our analysis has been restricted to the
627: noninteracting electron gas. To incorporate the effects of the
628: electron-electron interaction in the random phase approximation
629: one has to account for the self-consistent electric field
630: induced by the variations of the electron density. The potential for this field
631: $\phi_{sc}$ obeys the Poisson equation. In two dimensions the Fourier
632: transform of the Poisson equation has the form,
633: \begin{equation}
634: \label{poisson} e\phi_{sc} (\omega,{\bf q})= V_q \rho (\omega,{\bf
635: q}),
636: \end{equation}
637: where $V_q=2\pi e^2/q$ is the bare Coulomb propagator. The random
638: phase approximation (RPA) is then equivalent to the substitution
639: $\phi (\omega,{\bf q}) \to \phi_{sc} (\omega,{\bf q})+ \phi
640: (\omega,{\bf q})$ in the right-hand side of Eq. (\ref{linear}). It
641: is straightforward to see that the structure factor takes the
642: familiar RPA form,
643: \begin{equation}
644: \label{RPA} \chi_{RPA}(\omega,{\bf q})=\frac{\chi (\omega,{\bf
645: q})}{1-V_q \chi (\omega,{\bf q})}.
646: \end{equation}
647: The pole of this expression determines the plasmon spectrum
648: $\omega=\omega_q +i\gamma_q$, where $\omega_q^2 = v^2 \kappa q/2$,
649: with $\kappa=2\pi e^2 \nu$ standing for the static screening
650: radius. The plasmon linewidth is given by the imaginary part of
651: the bare structure factor at $\omega =\omega_q$,
652: \begin{equation}
653: \label{damping} \gamma_q = \frac{1}{2} V_q\omega_q |\Im
654: \chi(\omega_q,{q})|.
655: \end{equation}
656: For the plasmon to be an undamped excitation its frequency should
657: lie above the electron-hole continuum, $\omega_q > qv$, which
658: requires $\kappa
659: > 2q$. As was already pointed out in Ref. \onlinecite{MCE}
660: the plasmon acquires damping when $\omega_q \sim 2\Delta$. Since
661: $q \ll (q \kappa)^{1/2} \sim m\alpha$ at this range, the direct
662: transitions (\ref{imdirect}) make the principal contribution to
663: Eq.\ (\ref{damping}).
664:
665: {\it Impurity scattering}. The equations presented in this paper
666: assume ballistic electron motion. The absence of impurities allows
667: one to write kinetic equation as a closed set of equations, Eq.\
668: (\ref{tr}), for the density matrix integrated over the energy
669: variable $\epsilon$, [see Eq.\ (\ref{pauli})], i.e.\ at coinciding
670: times. In the presence of disorder the self-energy due to impurity
671: scattering should be added to the right-hand side of Eq.\
672: (\ref{set}). In general, since plain waves are no longer
673: eigenstates of the system with impurities, the equations for the
674: distribution function depending on the momentum ${\bf p}$ (and not
675: on the energy $\epsilon$) become not very convenient. More natural
676: (though more complicated) equations would result from integration
677: over $\xi_p$, similar to the usual spin-degenerate case\cite{RS}.
678: Such equations are beyond the scope of the present paper.
679:
680:
681: {\it Special case $\alpha=\pm \beta$.} Recently, Schliemann et.\
682: al.\ \cite{SEL} proposed a spin field-effect transistor based on a
683: particular tuning of the spin-orbit coupling constants such that
684: $\alpha=\beta$ (or $\alpha=-\beta$). This special system is
685: expected to preserve spin coherence even in the presence of
686: disorder. This is due to the fact that the spin eigenstates
687: (\ref{states}) are independent of the electron momenta, $\chi_{\bf
688: p} = \text{const}$, therefore a scalar impurity potential does not
689: result in the intersubband transitions. The same observation
690: holds for the structure factor. Since the matrix elements of the
691: density are identically zero for the transitions between different
692: subbands the second line of Eq. (\ref{chi}) is absent in this case
693: and the structure factor is intact by the presence of spin-orbit
694: interaction (up to higher order corrections).
695:
696: {\it Conclusions.} To summarize, we have derived transport
697: equations for the distribution function of a two-dimensional
698: electron gas with spin-orbit interaction of both the
699: Bychkov-Rashba and the Dresselhaus mechanisms. The distribution
700: function is a 2$\times$2-matrix in the spin space. General
701: expressions for the particle and energy currents and densities are
702: available in terms of the density $f_{\bf p}$ and spin ${\bf
703: g}_{\bf p}$ distribution functions. The obtained equations are
704: applied to the wave-packet propagation in a ballistic 2DEG and to
705: the calculation of the density-density correlation function
706: $\chi(\omega,q)$. We observe that for $q > m\alpha$ the structure
707: factor $\chi(\omega,q)$ is almost not affected by the spin-orbit
708: interaction, but it reveals new features when $q \ll m\alpha$ due
709: to the direct transitions between different spin-orbit subbands.
710:
711: Fruitful discussions with A.\ Andreev, K.\ Flensberg, L.
712: Glazman, and C.\ Marcus are gratefully appreciated. This material
713: is based on work supported by the NSF under grant PHY-01-17795 and
714: by the Defence Advanced Research Programs Agency (DARPA) under
715: Award No. MDA972-01-1-0024.
716:
717: \begin{thebibliography}{50}
718: \bibitem{ALS} {\it Semiconductor Spintronics and Quantum Computation},
719: editors, D.D. Awschalom, D. Loss, and N. Samarth, Springer,
720: Berlin, 2002.
721: \bibitem{DD} S. Datta and B. Das, Appl. Phys. Lett. {\bf 56}, 665
722: (1990).
723: \bibitem{WAB} S.A. Wolf, D.D. Awschalom, R.A. Buhtman, J.M.
724: Daughton, S. von Moln\'ar, M.L. Roukes, A.Y. Chtchelkanova, D.M.
725: Treger, Science {\bf 294}, 1488 (2001).
726: \bibitem{KNT} T. Koga, J. Nitta, H. Takayanagi, S. Datta, Phys.
727: Rev. Lett. {\bf 88}, 126601 (2002).
728: \bibitem{GBZ} M. Governale, D. Boese, U. Zulicke, C. Schroll,
729: Phys. Rev. B {\bf 65}, 140403 (2002).
730: \bibitem{KA} J.M. Kikkawa and D.D. Awschalom, Phys. Rev. Lett.
731: {\bf 80}, 4313 (1998).
732: \bibitem{FKR} R. Fiederling, M. Keim, G. Reuscher, W. Ossau, G.
733: Schmidt, A. Waag, L.W. Molenkamp, Nature (London) {\bf 402}, 790
734: (1999).
735: \bibitem{OYB} Y. Ohno, D.K. Young, B. Beschoten, F. Matsukura, H.
736: Ohno, D.D. Awschalom, Nature (London) {\bf 402}, 790 (1999).
737: \bibitem{ZRK} H.J. Zhu, M. Ramsteiner, H. Kostial, M. Wassermeier,
738: H.P. Sch\"onherr and K.H. Ploog, Phys. Rev. Lett. {\bf 87}, 016601
739: (2001).
740: \bibitem{RPS} M.Val\'in-Rodr\'igues, A. Puente and L. Serra,
741: cond-mat/0211694.
742: \bibitem{fer} C. Ciuti, J.P. McGuire, L.J. Sham,
743: cond-mat/0205651.
744: \bibitem{BR} Yu. Bychkov and E.I. Rashba, JETP Lett. {\bf 39}, 78
745: (1984).
746: \bibitem{NAT} J. Nitta, T. Akazaki, H. Takayanagi and T. Enoki,
747: Phys. Rev. Lett. {\bf 78} 1335 (1997).
748: \bibitem{Koga} T.\ Koga, J.\ Nitta, T.\ Akazaki, and H.
749: Takayanagi, Phys.\ Rev.\ Lett.\ {\bf 89}, 046801 (2002).
750: \bibitem{H} B.I. Halperin, A. Stern, Yu. Oreg, J.N.H.J. Cremers, J.A. Folk, and C.M.
751: Marcus, Phys. Rev. Lett. {\bf 86}, 2106 (2001).
752: \bibitem{AF} I.L. Aleiner and V.I. Fal'ko, Phys. Rev. Lett. 87, 256801 (2001)
753: \bibitem{M} D.M. Zumbuhl, J.B. Miller, C.M. Marcus, K. Campman, A.C.
754: Gossard, Phys.\ Rev.\ Lett.\ {\bf 89}, 276803 (2002).
755: \bibitem{MZM} J.B. Miller, D.M. Zumbuhl, C.M. Marcus, Y.B. Lyanda-Geller, D. Goldhaber-Gordon, K.
756: Campman, A.C. Gossard, Phys.\ Rev.\ Lett.\ {\bf 90}, 076807
757: (2003).
758: \bibitem{CR} G.-H.\ Chen and M.E.\ Raikh, Phys.\ Rev.\ B {\bf 59},
759: 5090 (1999).
760: \bibitem{MCE} L.I. Magarill, A.V. Chaplik and M.V. \'Entin, JETP
761: {\bf 92}, 153 (2001).
762: \bibitem{D} G. Dresselhaus, Phys. Rev. {\bf 100}, 580 (1955).
763: \bibitem{DK} M.I. Dyakonov and V.Y. Kachorovsii, Sov. Phys.
764: Semiconductors {\bf 20}, 110 (1986).
765: \bibitem{LP} E.M. Lifshitz and L.P. Pitaevskii, {\it Physical Kinetics},
766: (Pergamon, 1981).
767: \bibitem{HM} C.-M.\ Hu and T.\ Matsuyama, Phys.\ Rev.\ Lett. {\bf
768: 87}, 066803 (2001).
769: \bibitem{MH} T.\ Matsuyama, C.-M.\ Hu, D.\ Grundler, G.\ Meier,
770: and U.\ Merkt, Phys.\ Rev.\ B {\bf 65}, 155322 (2002).
771: \bibitem{LLF} M.H.\ Larsen, A.M.\ Lunde and K.\ Flensberg,
772: Phys.\ Rev.\ B {\bf 66}, 033304 (2002).
773: \bibitem{MSB} L.W. Molenkamp, G.\ Schmidt, and G.E. Bauer, Phys.\
774: Rev.\ B {\bf 64}, 121202 (2001).
775: \bibitem{Rashba} V.Ya.\ Kravchenko and E.I.\ Rashba, Phys.\ Rev.\ B {\bf 67}, 121310
776: (2003).
777: \bibitem{PN} D. Pines and P. Nozi\'eres, {\it The Theory of Quantum
778: Liquids} (Benjamin, Reading, Mass., 1966).
779: \bibitem{SEL} J.\ Schliemann, J.C.\ Egues and D.\ Loss, Phys.\ Rev.\ Lett.\ {\bf 90}, 146801 (2003).
780: \bibitem{RS} J.\ Rammer and H.\ Smith,
781: Rev. Mod. Phys. {\bf 58}, 323 (1986).
782:
783: \end{thebibliography}
784: \end{document}
785: